What Olbrain is, what you can build on it, how it works, the models it uses, where data lives, how security and audit work, and what it costs.
Olbrain is the Agent OS for Enterprises — the platform where you build, run, and govern enterprise AI agent systems. Its lead orchestrator is an AI agent that builds the agent systems for you: you describe the goal in natural language and the orchestrator directs the specialists from discovery through build and deployment. Olbrain is live at studio.olbrain.com.
Five kinds of agent on one orchestration layer: Concierge (conversational), Operator (workflow), Analyst (research), Orchestrator, and Designate (role-holding — humans supervise). Three runtimes are live and shipping today — conversational, workflow, and research. Typical agents include customer support, KYC and onboarding, collections and NACH, TDS verification, and document processing.
On Olbrain, agents author the agents: the lead orchestrator and its specialist sub-agents do the discovery, requirements, build, and instrumentation, so your team brings the goal and stays on review and the hard edge cases. Every agent runs as a self-identity under the Agency Protocol, and every action is captured in an immutable audit log with signed, PII-free receipts. Olbrain builds and owns its entire agent stack, including its own orchestration layer, and is India-hosted.
The lead orchestrator is an AI agent inside Olbrain that builds your entire enterprise agent systems. It has full operator-level access to the platform; you describe the goal in natural language, and it directs a team of specialist sub-agents through discovery, build, and deployment. Once built, agents run and are governed within the same platform.
The Agency Protocol is Olbrain’s mechanism for giving each agent a self-identity — one accountable, exclusive actor — so that every decision and action is attributable to it and the enterprise can answer for what its AI does.
Olbrain is model-agnostic. It calls leading language models directly and can swap models without changing the agent, so the platform adopts the newest, most capable, or most compliant model as it becomes available. Language-model costs are passed through at cost.
Olbrain customer data is stored and processed in India — Google Cloud, Mumbai region (asia-south1). The architecture is DPDP-aligned and built for RBI data-residency requirements. PII is tokenized before any payload reaches a language model.
Security is built in: an immutable, append-only audit log, a signed PII-free receipt for every action, PII tokenization before any model call, tenant isolation, per-tenant key management, and role-based access control. Customer and enterprise data is never used to train base models, fine-tunes, or evaluations. SOC 2 and ISO 27001 programs are in progress. Full, honestly status-labelled detail is on the Trust page.
Every agent action is written to an immutable, append-only audit log and emits a signed, PII-free receipt. Any action can be replayed and issued as an audit packet on demand for InfoSec, Risk, or a regulator. Accountability stays with the enterprise; Olbrain provides the infrastructure that makes the enterprise’s AI something it can attribute, verify, and answer for.
₹20,000 per agent per month (includes build, hosting, platform updates, and support), plus ₹2 per AI message for conversational usage, ₹1 per step for workflow usage, and from ₹5,000 per research report. Language-model costs are included in the price. Billing is monthly with no lock-in and no minimum.
Enterprises that need to build and govern agent systems under regulatory scrutiny — beginning with Indian regulated finance (NBFCs, banks, insurers, asset managers, fintechs) — where auditable identity, attributable decisions, and data residency matter.
Yes. Ten agents run in production — five build and run customer agent systems, and five run Olbrain itself (engineering documentation, production debugging, DevOps, hiring, and sales). Olbrain is in active UAT with a leading Indian NBFC of roughly $1B AUM: four agents in UAT, one at BRD, and sixty-plus workflows in scope.
No. Olbrain builds and owns its entire agent stack, including its own orchestration layer. Language models are called directly and are swappable, so the platform can adopt new models the day they ship.
Agents reach external systems through tools — implemented as Model Context Protocol (MCP) tool servers — and connectors.
Olbrain Labs Private Limited, headquartered in Gurugram, India (CIN U62013HR2024PTC123898), co-founded by Alok Gotam (Founder & CEO) and Nishant Singh (Founder & CTO). The Olbrain mission has been continuous since February 2017; the current operating entity was registered in August 2024. More in the press kit.